The following bug has been logged on the website:
Bug reference: 17296
Logged by: Anna
Email address: podkina@gmail.com
PostgreSQL version: 12.8
Operating system: Oracle Linux Server release 8.4
Description:
Hello!
I've met the problem with the cluster by patroni.
Cluster configuration: 3 nodes, etcd downstairs.
I tried to set up the Standby cluster through the replication slot as I did
it before on the same version of patroni/PostgreSQL, but different OS and
hardware environment.
(not sure it's important)
Anyway, I create the replication slot by the command (psql, locally on the
host):
SELECT pg_create_physical_replication_slot('patroni12_standby', true);
Immediately after I see my slot in pg_replication_slots
Then in a few seconds, there is no slot.
On the other side (on Standby PostgreSQL) I see in a log:
2021-11-22 13:15:35 MSK [96831]:
user=replicator,db=[unknown],app=[unknown],client=*** LOG: replication
connection authorized: user=replicator application_name=***.io
2021-11-22 13:15:35 MSK [96831]:
user=replicator,db=[unknown],app=***.io,client=*** ERROR: replication slot
"patroni12_standby" does not exist
2021-11-22 13:15:35 MSK [96831]:
user=replicator,db=[unknown],app=***.io,client=*** STATEMENT:
START_REPLICATION SLOT "patroni12_standby" 34A9/EA000000 TIMELINE 8
2021-11-22 13:15:35 MSK [96831]:
user=replicator,db=[unknown],app=***.io,client=*** LOG: disconnection:
session time: 0:00:00.049 user=replicator database= host=*** port=35772
### here I create replication slot on master node
2021-11-22 13:15:56 MSK [300984]:
user=replicator,db=[unknown],app=[unknown],client=*** LOG: replication
connection authorized: user=replicator application_name=***.io
2021-11-22 13:15:56 MSK [300984]:
user=replicator,db=[unknown],app=***.io,client=*** ERROR: requested WAL
segment 00000008000034A9000000EA has already been removed
2021-11-22 13:15:56 MSK [300984]:
user=replicator,db=[unknown],app=***.io,client=*** STATEMENT:
START_REPLICATION SLOT "patroni12_standby" 34A9/EA000000 TIMELINE 8
2021-11-22 13:15:56 MSK [300984]:
user=replicator,db=[unknown],app=***.io,client=*** LOG: disconnection:
session time: 0:00:00.056 user=replicator database= host=*** port=54038
### The WAL segment error doesn't matter, but It show that the slot exists.
Then I don't do anything and next log/error about "slot does not exist"
again
2021-11-22 13:16:01 MSK [300998]:
user=replicator,db=[unknown],app=[unknown],client=*** LOG: replication
connection authorized: user=replicator application_name=***.io
2021-11-22 13:16:01 MSK [300998]:
user=replicator,db=[unknown],app=***.io,client=*** ERROR: replication slot
"patroni12_standby" does not exist
Also, I tried to create a slot without a readiness standby node but the
behavior was the same. The slot exists for short time and then it's
disappearing.
All versions:
OS: Oracle Linux Server release 8.4
PG: PostgreSQL 12.8 on x86_64-pc-linux-gnu, compiled by gcc (GCC) 8.4.1
20200928 (Red Hat 8.4.1-1), 64-bit
Orchestrator: patroni 2.1.1 / patronictl version 2.1.1 / etcd 3.4.16
PGconfiguration:
autovacuum_analyze_scale_factor: 0.01
autovacuum_vacuum_scale_factor: 0.01
checkpoint_completion_target: 0.7
default_statistics_target: 100
effective_cache_size: 200GB
effective_io_concurrency: 250
hot_standby: 'on'
hot_standby_feedback: 'on'
log_autovacuum_min_duration: 0
log_checkpoints: true
log_connections: true
log_disconnections: true
log_line_prefix: '%t [%p]: user=%u,db=%d,app=%a,client=%h '
log_lock_waits: true
log_min_duration_statement: 1000
log_statement: ddl
log_temp_files: 100000
maintenance_work_mem: 4GB
max_connections: 1000
max_parallel_maintenance_workers: 4
max_parallel_workers: 12
max_parallel_workers_per_gather: 4
max_replication_slots: 10
max_wal_senders: 10
max_wal_size: 4GB
max_worker_processes: 128
min_wal_size: 1GB
random_page_cost: 1.0
seq_page_cost: 1.0
shared_buffers: 156GB
wal_buffers: 16MB
wal_keep_segments: 8
wal_level: replica
work_mem: 128MB
Ready to provide you with any extended information.
I hope that's not a bug, but I couldn't find anything about it on net
Best regards,
Anna